Impact of genotype x environment interaction and selection history on genomic prediction in maize (Zea mays L.)

Authors: COSTA, MARTIN - University Of Wisconsin; Holland, Jim - Jim; DE LEON, NATALIA - University Of Wisconsin; KAEPPLER, SHAWN - University Of Wisconsin

Submitted to: Crop Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/3/2024
Publication Date: 10/15/2024
Citation: Costa, M., Holland, J.B., De Leon, N., Kaeppler, S.M. 2024. Impact of genotype x environment interaction and selection history on genomic prediction in maize (Zea mays L.) . Crop Science. 64(6):3293-3310. https://doi.org/10.1002/csc2.21379.
DOI: https://doi.org/10.1002/csc2.21379

Interpretive Summary: Corn breeders seek to develop new hybrid varieties with improved yield and agronomic performance across a range of environments. It has been uncertain whether selecting for optimal performance in testing settings leads to reduced stability of yield across environments. In this study we compared three different corn populations representing different time periods of improvement for their yield performance across a wide range of environments. The maize Genomes 2 Fields project provides an ideal platform for testing these ideas, as it tests common sets of corn hybrids across 50 or more environments in the USA every two-year period. We found that the population representing the most recent period of breeding had the greatest yield stability across environments, suggesting that commercial breeders have successfully improved both mean yields and yield stability across environments.

Technical Abstract: Breeders made remarkable progress in improving productivity and stability of cultivars. Breeding progress relies on selecting favorable alleles for performance and stability to produce productive varieties across diverse environments. In this study, we analyzed the Genomes to Fields Initiative 2018–2019 genotype by environment interaction (G × E) dataset, focusing on three populations of double haploid (DH) lines derived from crossing inbrexpired Plant Variety Protection (ex-PVP) inbred line PHW65 with inbred lines PHN11, Mo44, and MoG. PHW65 is an Iodent/Lancaster-type inbred; PHN11 is an Iodent type ex-PVP line; Mo44 is a tropical-derived inbred; and MoG is an agronomically poor line derived from the variety Mastadon. Hybrids were produced by crossing the resulting DHs with Stiff Stalk testers PHT69 and LH195. The study's objective was to determine the donor inbreds' relative value and understand the impact of selection history on genomic prediction. We conducted a two-stage analysis to compare hybrid performance and G × E variance of the populations. G × E variance for yield was significantly lower in the PHW65 × PHN11 population relative to the PHW65 × MoG population. The reduced G × E variance of the PHN11 population led to increased indirect prediction accuracy (when training and testing data are drawn from the same population but different environments). In cross-validation, the PHN11 population had the greatest indirect prediction accuracy 45% of the time, followed by the Mo44 population (30%) and the MoG population (25%). Results demonstrate that prediction accuracy was greater in the population with the longest history of selection for favorable alleles (PHN11), contributing to greater yield stability.